CN105303560A - Robot laser scanning welding seam tracking system calibration method - Google Patents

Abstract

The invention discloses a robot laser scanning welding seam tracking system calibration method. A video camera in a sensor is calibrated based on a plane template firstly. The video camera is enabled to acquire at least three different positions by freely moving the sensor. The internal parameters of the video camera are solved according to the principle of least squares. Then image coordinates of an infinite point on a laser plane projected in the video camera are solved, and structured light is calibrated. Finally the calibrated sensor is fixed at the tail end of a manipulator via a flange plate. The same target point in space is detected by using the sensor through changing the attitude of a robot, and a conversion matrix of a video camera coordinate system and a manipulator tail end coordinate system is solved. A robot laser scanning welding seam tracking system acts as the object of study, and the parameters of the structured light are calibrated by adopting the video camera calibration method based on the plane template and the image coordinates of the infinite point on the laser plane projected in the video camera.

Description

Robotic laser's scanning type weld seam tracking system calibrating method

Technical field

The present invention relates to the demarcation of robotic laser's scanning type weld seam tracking system, comprise the demarcation of laser scan type weld seam tracking sensor and the demarcation of Robot Hand-eye.

Background technology

Demarcation and the Robotic Hand-Eye Calibration of laser scan type weld seam tracking sensor are depended in the accurate tracking of laser scan type seam tracking system butt welded seam.The demarcation of sensor comprises camera calibration and calibration, and the Technical comparing of camera calibration is ripe.For the demarcation of laser structure light, the method used the earliest has profile of tooth standardization and wire drawing standardization, and these methods utilize specific demarcation to harrow and produce calibration point in structured light plane, and utilize auxiliary measuring instrument to obtain the coordinate of its calibration point in structured light plane.These two kinds of methods need accurate utility appliance.The classic method of Robotic Hand-Eye Calibration has and utilizes the instrument of high precision three-dimensional coordinates measurement instrument to sensor holders to measure, and the method is all not suitable for the hand and eye calibrating that end installs the robot of vision sensor.

Summary of the invention

For the deficiency of existing scaling method, the invention provides a kind of robotic laser's scanning type weld seam tracking system calibrating method.

The step of the technical solution used in the present invention is as follows:

Step 1. adopts demarcates the video camera in sensor based on plane template, by moving freely sensor, makes camera acquisition arrive the template image of at least three diverse locations (relative to video camera).The inner parameter of video camera is solved according to the principle of least square.

Step 2. asks for the image coordinate that on laser plane, infinite point a bit projects in video camera, demarcates structured light.

The sensor demarcated is fixed on arm end by ring flange by step 3., uses same impact point in sensor detection space, solve the transition matrix of camera coordinate system and arm end coordinate system by the attitude changing robot.

Beneficial effect of the present invention: the present invention with robotic laser's scanning type weld seam tracking system for research object, adopt the image coordinate a bit projected in video camera based on infinite point on the camera marking method of plane template and laser plane, structured light parameter is demarcated.And the logical attitude changing robot is measured impact point same in experimental field, calculates the trick matrix of this system fast.

Accompanying drawing explanation

Fig. 1 is robotic laser's seam tracking system;

Fig. 2 is laser scan type weld seam tracking sensor structure;

Fig. 3 is sensor measurement model;

Fig. 4 demarcates target;

Fig. 5 is that fixed point becomes pose measurement.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention will be further described.

(1) robotic laser's seam tracking system is as shown in Figure 1,

Robotic laser scans seam tracking system primarily of six-DOF robot and laser scan type weld seam tracking sensor composition, comprises robot 1, laser seam tracking sensor 2, word structured light 3 and face of weld 4.The coordinate system that this system relates to has as follows:

In figure, { R} is the base coordinate system be connected with robot base; { E} is arm end coordinate system; { C} is camera coordinate system.

(2) hardware configuration of sensor and principle of work thereof

As shown in Figure 2, laser scan type weld seam tracking sensor is made up of laser instrument 6, filter plate 8, industrial camera 5, connector 7 and spatter shield 9 etc.The perforate of sensor cabinet top, for power lead and the signal wire extraction of video camera, laser instrument.Sensor probe is connected with computing machine by signal wire.The spatter shield that sensor bottom of shell is made up of organic glass covers, the splashing produced when can stop welding, protection video camera.Semiconductor laser tilts to be fixed in sensor probe casing.The laser beam that one word laser instrument sends, oblique illumination, to the commissure on welding work pieces surface, forms laser stripe.Camera acquisition to the laser stripe image comprising weld information, be sent to computing machine by signal wire.

As shown in Figure 3, the measurement model of this sensor employing.In figure, π is the laser plane of laser instrument projection; O _c-X _cy _cz _cfor the camera coordinate system set up for initial point with optical axis and lens intersection point; O-XY is image coordinate system in video camera.There is 1 P (x in laser plane _c, y _c, z _c), be projected as p (x, y) at the plane of delineation.Following relation is obtained according to triangulation principle:

Wherein, f is focal length of camera; B is laser plane and X _caxle intersection point, to the distance of optical axis, is called that base is long; θ is X _cangle with laser plane, is called projectional angle.The camera coordinates of impact point and the transformational relation of image coordinate are:

If focal length of camera f, projectional angle θ, the long b of base are known in formula, then from formula (2), in three dimensions, any point all can with the some one_to_one corresponding in two dimensional image.The Internal system parameters solving f, θ, b and laser sensor system is demarcated.Wherein focal distance f is obtained by camera calibration.

(1) demarcation of sensor

When ascending a height to enjoy a distant view, the ground level of infinite point is straight line, and equally, two parallel lines at infinity meet at a bit.As shown in Figure 4, will demarcate target horizontal positioned, and regulate the position of sensor, and make the optical axis of video camera perpendicular to demarcation target, laser plane and two parallel lines meet at A, B 2 point, and control video camera moves along optical axis direction, obtains a series of intersection point A _i, B _i(i=1,2 ... n), it is projected as a in video camera _i, b _i(i=1,2 ... n).A _i, B _ilaser plane forms two parallel lines, and it at infinity meets at a bit.By a _iwith b _ifit to straight line respectively, on the intersection point of two straight lines and laser plane, infinite point a bit projects in video camera.

There is 1 P (x in laser plane _c, y _c, z _c) be p (x, y) in the picture point of the plane of delineation.According to triangulation principle, can obtain:

The P that sets up an office be on laser plane infinite point a bit, i.e. (x _c, y _c, z _c) → ∞, z _c→ ∞, can learn fcot θ-x _∞→ 0 and get final product:

cotθ＝x _∞/f⑷

Wherein, x _∞for the image coordinate that infinite point on laser plane a bit projects in video camera.

If will write as total differential form, can be obtained:

If make Δ x=0 namely can obtain:

(6) can be obtained by formula:

Wherein, x is the image coordinate of current goal point; If Δ y can be controlled _c, and obtaining corresponding Δ y by image procossing, then the long b of the base of structured light can determine.

(2) Robotic Hand-Eye Calibration

As shown in Figure 1, T _eRthat robot end { relative to pedestal mark, { can obtain from the Eulerian angle the controller of robot, comprise rotation matrix R E} by the transition matrix of R} _eRwith translation vector t _eR, T _cEfor camera coordinate system, { relative to arm end coordinate system, { transition matrix of E} comprises rotation matrix R to C} _cEwith translation vector t _cE.Because sensor is fixed in arm end, therefore T _cEdo not change with the motion of mechanical arm.

As shown in Figure 5, by changing the same impact point M of attitude repeated detection of robot, robot is under i-th pose, and the coordinate of impact point M under camera coordinate system is M _i ^c, the coordinate under basis coordinates system of robot is M ^r, there is following transformational relation:

Wherein, T _eRifor robot i-th pose, arm end, relative to the transition matrix of basis coordinates system, comprises rotation matrix R _eRiwith translation vector t _eRi.

(9) formula is launched to obtain:

Control mechanical arm, can obtain just to repetitive measurement impact point M:

1) rotation matrix R _cEsolve

If in motion process, the attitude of robot remains unchanged is R _eR1=R _eR2=...=R _eRn=R ₀.

Due to R ₀be orthogonal matrix, namely have can be obtained by repetitive measurement:

Order then (12) formula can be written as R _cEa ₁=b ₁matrix equation, solve by decomposition of singular matrix method (SVD) Q, namely

R _CE＝VU ^T⒀

Wherein V and U is respectively matrix right singular matrix and left singular matrix.

2) translation vector t _cE

Can obtain according to formula (10):

The multiple attitude of control mechanical arm is carried out detection to impact point and can be obtained type as A ₂t _cE=b ₂

Wherein A ₂=[R _eR2-R _eR1r _eR3-R _eR1r _eRn-R _eR1], b ₂=[R _eR1r _cEm ₁ ^c-

R _ER2R _CEM ₂ ^C+t _ER1-t _ER2R _ER1R _CEM ₁ ^C-R _ER3R _CEM ₃ ^C+t _ER1-t _ER3…

R _ER1R _CEM ₁ ^C-R _ERnR _CEM _n ^C+t _ER1-t _ERn]。

Claims (1)

1. robotic laser's scanning type weld seam tracking system calibrating method, is characterized in that the method comprises the following steps:

Step 1. adopts demarcates the video camera in sensor based on plane template, by moving freely sensor, making camera acquisition arrive the template image of at least three diverse locations, solving the inner parameter of video camera according to the principle of least square;

Step 2. asks for the image coordinate that on laser plane, infinite point a bit projects in video camera, demarcates structured light, specifically:

To demarcate target horizontal positioned, and regulate the position of sensor, and make the optical axis of video camera perpendicular to demarcation target, laser plane and two parallel lines meet at A, B 2 point, and control video camera moves along optical axis direction, obtains a series of intersection point A _i, B _i, i=1,2 ... n, it is projected as a in video camera _i, b _i; Intersection point A _i, B _ilaser plane forms two parallel lines, and it at infinity meets at a bit; By a _iwith b _ifit to straight line respectively, on the intersection point of two straight lines and laser plane, infinite point a bit projects in video camera;

There is 1 P (x in laser plane _c, y _c, z _c) be p (x, y) in the picture point of the plane of delineation; According to triangulation principle:

$\left\{\begin{array}{c}{x}_c=\frac{b\·x}{f\cot \mathrm{\θ}-x}\\ y_c=\frac{b\·y}{f\·\cot \mathrm{\θ}-x}\\ z_c=\frac{b\·f}{f\·\cot \mathrm{\θ}-x}\end{array}\right.---\left(3\right)$

The P that sets up an office be on laser plane infinite point a bit, i.e. (x _c, y _c, z _c) → ∞, z _c→ ∞, can learn fcot θ-x _∞→ 0 and get final product:

cotθ＝x _∞/f⑷

Wherein, x _∞for the image coordinate that infinite point on laser plane a bit projects in video camera; If will write as total differential form, can be obtained:

${\mathrm{\Δy}}_c=\frac{\∂y_c}{\∂y}+\frac{\∂y_c}{\∂x}=\frac{b}{f\cot \mathrm{\θ}-x}\·\mathrm{\Δ}y+\frac{\∂y_c}{\∂x}.\mathrm{\Δ}x---\left(5\right)$

If make Δ x=0 namely can obtain:

${\mathrm{\Δy}}_c=\frac{b}{f\cot \mathrm{\θ}-x}\·\mathrm{\Δ}y---\left(6\right)$

(6) can be obtained by formula:

$b=\frac{f\cot \mathrm{\θ}-x}{\mathrm{\Δ}y}\·{\mathrm{\Δy}}_c---\left(7\right)$

Wherein, x is the image coordinate of current goal point; If Δ y can be controlled _c, and obtaining corresponding Δ y by image procossing, then the long b of the base of structured light can determine;

The sensor demarcated is fixed on arm end by ring flange by step 3., uses same impact point in sensor detection space, solve the transition matrix of camera coordinate system and arm end coordinate system by the attitude changing robot, specifically:

By changing the same impact point M of attitude repeated detection of robot, robot is under i-th pose, and the coordinate of impact point M under camera coordinate system is M _i ^c, the coordinate under basis coordinates system of robot is M ^r, there is following transformational relation:

$M^R=T_{ERi}{T}_{CE}{M}_i^C---\left(8\right)$

$\left[\begin{array}{c}{M}^R\\ 1\end{array}\right]=\left[\begin{array}{cc}{R}_{ERi}& t_{ERi}\\ 0& 1\end{array}\right]\left[\begin{array}{cc}{R}_{CE}& t_{CE}\\ 0& 1\end{array}\right]\left[\begin{array}{c}{M}_i^c\\ 1\end{array}\right]---\left(9\right)$

Wherein, T _eRifor robot i-th pose, arm end, relative to the transition matrix of basis coordinates system, comprises rotation matrix R _eRiwith translation vector t _eRi;

(9) formula is launched to obtain:

$M^R=R_{ERi}{R}_{CE}{M}_i^c+R_{ERi}{t}_{CE}+t_{ERi}---\left(10\right)$

Control mechanical arm, can obtain just to repetitive measurement impact point M:

$\left\{\begin{array}{c}{M}^R=R_{ER1}{R}_{CE}{M}_1^c+R_{ER1}{t}_{CE}+t_{ER1}\\ M^R=R_{ER2}{R}_{CE}{M}_2^c+R_{ER2}{t}_{CE}+t_{ER2}\\ \begin{array}{c}.\\ .\\ .\end{array}\\ M^R=R_{ERn}{R}_{CE}{M}_n^c+R_{ERn}{t}_{CE}+t_{ERn}\end{array}\right.---\left(11\right)$

1) rotation matrix R _cEsolve

If in motion process, the attitude of robot remains unchanged is R _eR1=R _eR2=...=R _eRn=R ₀;

Due to R ₀be orthogonal matrix, namely have can be obtained by repetitive measurement:

$\begin{array}{c}{R}_{CE}\left[\begin{array}{cccc}{M}_2^c-M_1^c& M_3^c-M_1^c& ...& M_n^c-M_1^c\end{array}\right]\\ =R_0^T\left[\begin{array}{cccc}{t}_{ER1}-t_{ER2}& t_{ER1}-t_{ER3}& ...& t_{ER1}-t_{ERn}\end{array}\right]\end{array}---\left(12\right)$

Order $A_1=\left[\begin{array}{cccc}{M}_2^c-M_1^c& M_3^c-M_1^c& ...& M_n^c-M_1^c\end{array}\right];$

$b_1=R_0^T\left[\begin{array}{cccc}{t}_{ER1}-t_{ER2}& t_{ER1}-t_{ER3}& ...& t_{ER1}-t_{ERn}\end{array}\right]$

Then (12) formula can be written as R _cEa ₁=b ₁matrix equation, solve by decomposition of singular matrix method Q, namely

R _CE＝VU ^T⒀

Wherein V and U is respectively matrix right singular matrix and left singular matrix;

2) translation vector t _cE

Can obtain according to formula (10):

$\begin{array}{c}\left(R_{ER2}-R_{ER1}\right)t_{CE}\\ =R_{ER1}{R}_{CE}{M}_1^c-R_{ER2}{R}_{CE}{M}_2^c+t_{ER1}-t_{ER2}\end{array}---\left(14\right)$

The multiple attitude of control mechanical arm is carried out detection to impact point and can be obtained type as A ₂t _cE=b ₂

$t_{CE}={\left(A_2^T{A}_2\right)}^{-1}{A}_2^T{b}_2---\left(15\right)$

Wherein A ₂=[R _eR2-R _eR1r _eR3-R _eR1r _eRn-R _eR1],

b ₂＝[R _ER1R _CEM ₁ ^C-R _ER2R _CEM ₂ ^C+t _ER1-t _ER2R _ER1R _CEM ₁ ^C-R _ER3R _CEM ₃ ^C+t _ER1-t _ER3…R _ER1R _CEM ₁ ^C-R _ERnR _CEM _n ^C+t _ER1-t _ERn]。